An epitome-based evolutionary algorithm with behavior division for multimodal optimizations

In this paper, a novel evolutionary algorithm (EA) with two groups is presented based on the mimicry of a two-group team for a specific objective. The operations of exploration and epitome-based learning behaviors are properly defined. By means of the inherited generation of new individual and the replacement rules of the team, the behavior division between the elite group and the plain group is established, which make the algorithm have the potential for adaptive local, global and directive search. The conflict between the successful global search and the fast convergence in some other algorithms can be obviously mitigated in this algorithm. It can be shown by the comparisons that the presented algorithm is statistically superior to the genetic algorithm and particle swarm optimization in both global optimization and computational cost for multimodal optimization.